Research Project B05

The aim of this project is to characterize fractures and fracture networks based on a combination of physical experiments, image-based methods, and numerical simulations. The focus of all the investigations lies on the fundamental understanding of hydromechanical coupling effects of (deformable) fluid-saturated fractures embedded in a deformable porous material. Project B05 aims at providing the scientific community with detailed information and further insight into i) the development and evolution of preferential flow paths in fracture networks, ii) the role of asperities in the non-linear effective hydromechanical coupling between fractures, iii) the possibility of back-analyze  the fracture properties in the case of non-linear hydromechanical coupling, iv) the role of leak-off in crystalline rocks, and v) whether the same concepts can be applied on fractures saturated with more than one fluid phase.

For the investigation of complex fracture networks, we quenched (gradual increase and sudden decrease of temperature) a Bianco Carrara marble. The corresponding samples were characterized with standard techniques for their effective properties before and after treatment. Additionally, the topology of the fracture network was investigated with μ-XRCT. For the effective characterization of the fracture network, the segmentation the μ-XRCT data had to be performed. In order to compare the segmentation efficiency, we adopted three conventional segmentation methods and two machine learning segmentation methods. For the conventional segmentation methods, the application of a data noise reduction technique before segmentation was essential for the success of the process. That was not necessary in the machine learning schemes, which eventually saves computation time. However, given that the training of the machine learning model required excessive memory, the images were diced into small tiles in order to avoid a memory limitation issue. The U-net machine learning scheme was proven to be the most effective approach to segment the μ-XRCT data.

The in-house developed Digital Rock Physics (DRP) workflow was also employed to complement the experimental approach with XRCT-based Stokes simulations for large REVs (up to1000³ voxels. An explicit high performance Stokes solver has been developed and it is currently being tested for voxel-based geometries XRCT scans.

We also focused on the hydro-mechanical coupling in a fracture. A numerical investigation was performed which was able to reproduce a non-linear relation between flux and pressure caused by a volume change, under a harmonic pressure-/flow stimulation of a single fracture under confining stress. Under the same conditions, a benchmark experiment was conducted on an aluminum sample with rough fracture surfaces. This artificial fracture was stimulated with a harmonic flux of a fluid via an induced bore hole and the corresponding induced pressure was measured. As a way to show the non-linearity in the relation between flux and pressure under a harmonic stimulation, we present the measured corresponding pressure in the frequency domain. In both the numerical and experimental investigation, overtones were observed which imply the non-linear response of the system.

However, an artificially induced fracture may hold different properties such as storativity, diffusivity and contact stiffness, from natural fractures. For further realistic investigation of the harmonic stimulation on a single fracture, we introduced a single fracture in a sandstone sample by hydraulic pressure. Performing the harmonic experiment on this realistic sample is ongoing work.

Microfluidic investigations have also been performed in order to understand the hydromechanical interaction of “soft” fractures embedded in Poly-di-methyl-siloxane (PDMS) We investigated the intrinsic pressure diffusion time of single fractures and fracture networks with respect to the fluid’s viscosity and the elastic properties of PDMS.

Project B05, in the process of addressing the open questions stated in the beginning, the following steps will be made.

Fractured rock, microfluidic, and numerical investigations will take place in the attempt to address the objectives of the project. This includes μ-XRCT investigations of natural samples both in-the-house, and in a synchrotron environment for fast, time-resolved 3D characterization. Additionally, 2D, time- and space-resolved microfluidic investigations will take place, both with “soft” and “hard” micromodels. Complementary to the experimental investigations, numerical investigations will be performed, also in a supplementary manner for the flow domains that cannot be materialized with microfluidics due to technical limitations.

With our external partners from the Lawrence Berkeley National Laboratory, Dr. Jens Birkholzer and Dr. Seiji Nakagawa, we will perform transient/harmonic rock physics experiments under lithostatic stress states. Further, we are aiming at the analysis, clustering and visualization of high-resolution XRCT data obtained from static and, especially, dynamic synchrotron investigations, in collaboration with our external partners at the University of Manchester (Dr. Vahid Joekar-Niasar).